Individual Determinants of Ethnic Identification

Texte intégral

(1)DOCUMENT DE TRAVAIL. DT/2011-06. Individual Determinants of Ethnic Identification. Thomas BOSSUROY. UMR DIAL 225. Place du Maréchal de Lattre de Tassigny 75775 • Paris Cedex 16 •Tél. (33) 01 44 05 45 42 • Fax (33) 01 44 05 45 45 • 4, rue d’Enghien • 75010 Paris • Tél. (33) 01 53 24 14 50 • Fax (33) 01 53 24 14 51 E-mail : dial@dial.prd.fr • Site : www.dial.prd.fr.

(2) INDIVIDUAL DETERMINANTS OF ETHNIC IDENTIFICATION 1 Thomas Bossuroy SALDRU, University of Cape Town, South Africa UMR DIAL-Paris-Dauphine thomas.bossuroy@gmail.com Document de travail UMR DIAL Avril 2011 Abstract This paper examines the individual incentives to identify to one's ethnic group rather than to the nation, based on large sample surveys representative of seven capitals of West-African countries. Three main driving forces stand out. First, we show that education brings down ethnic salience at the individual level, contrary to claims in the literature that it stimulates ethnic-based competition within the elite. Second, ethnic identification is more frequent among those left out of the job market, like uneducated unemployed or informal workers who seek a new or better job, and is raised by the share of the individual's ethnic group integrated on the job market. Third, ethnic identification is higher among migrants, and again positively correlated to the share of the migrant's ethnic group that is employed. These results point to the role of ethnic groups as solidarity networks for individuals deprived of access to good jobs. Ethnic identification may reflect an investment in a specific kind of individual social capital with classical economic properties. The less social capital people have initially -due to poor education or recent migration for instance- and the more they need it -to escape unemployment or bad jobs-, the more they use ethnic ties to climb the social ladder. In that sense, ethnicity appears as a substitute to the formal means of social rise, and the initial deprivation of the latter fosters individual ethnic salience. Keywords: Ethnicity, Identity, Social capital, Networks, Africa Résumé Ce papier examine les incitations individuelles à s’identifier à son groupe ethnique plutôt qu’à sa nation, en analysant des enquêtes à large échantillon représentatives des capitales de sept pays ouestafricains. Trois principaux facteurs apparaissent. Premièrement, l’éducation affaiblit l’identification ethnique au niveau individuel, plutôt que de stimuler une compétition à base ethnique au sein de l’élite comme évoqué parfois dans la littérature. Deuxièmement, l’identification ethnique est plus fréquente chez les individus exclus du marché du travail, comme les chômeurs non éduqués ou les travailleurs informels qui cherchent un emploi, et augmente avec la part du groupe ethnique de l’individu intégrée sur le marché du travail. Troisièmement, l’identification ethnique est plus forte chez les migrants, et est corrélée positivement avec la part du groupe ethnique du migrant qui est employée. Ces résultats révèlent le rôle des groupes ethniques comme réseaux de solidarité pour les individus privés d’accès aux emplois formels. L’identification ethnique semble refléter un investissement dans une forme particulière de capital social aux propriétés économiques classiques. Moins les individus possèdent de capital social initial (en raison d’une faible éducation ou d’une migration récente par exemple), et plus ils en ont besoin (pour sortir du chômage ou accéder à un emploi protégé), plus ils utilisent le lien ethnique pour gravir l’échelle sociale. L’ethnicité venant ainsi se substituer aux modes formels d’ascension sociale, la privation de ces derniers renforce l’importance des attachements ethniques individuels. Mots Clés : Ethnicité, Identité, Capital social, Réseaux, Afrique. JEL Codes : A13, A14, D74, O17 1. Most of the research was done as a Ph.D. candidate at the Paris School of Economics, DIAL, Paris. I wish to thank Denis Cogneau, Eliana La Ferrara, Ted Miguel and Joel Selway for their help and comments on various drafts. Financial support from the Agence Française de Développement is gratefully acknowledged..

(3) 1 Introduction The understanding of what an ethnic group is and what ethnic identication means evolved a lot in the last fty years. The conceptual shift was initiated by anthropologists in the 1960s, but only impacted economic works very recently. In the 1950s, anthropologists were seeking to dene and identify ethnic groups, conceived as essential, unchanging core entities. This required to nd out their most critical and distinctive features, and led to attempts to establish robust taxonomies on that basis (see for example the works by Nadel (1951), Murdock (1957) or Naroll (1964)). This range of works led to the production of Human Relations Area Files, and Ethnographical Atlas. The most famous of them, Atlas Narodov Mira (1964), made in the Soviet Union, constitutes the basic material for the computation of the widely used ethno-linguistic fractionalization (ELF) indicator. This essentialist approach was yet to be rapidly questioned. In the 1960s, anthropological eld works strongly suggested that ethnic lines are much evolutive and that their salience varies along with political, social or economic determinants. On the basis of his works on the Kachins in Burma, Leach (1964) went as far as denouncing ethnic classications as "tabulated nonsense". Barth (1969) proposed a groundbreaking theory on the boundaries of ethnic groups which underlined the subjective dimension of ethnic identication and the importance of interpenetration and interdependency of ethnic groups. These works do not deny the existence and importance of ethnic groups. But they show how evolutive and socially constructed they may be. These theories were even reinforced by various political demands for recognition throughout the world (Glazer and Moynihan 1963), which highlighted how the reference to ethnic membership may vary across time. Indeed, contrary to the predictions of evolutionist theories that they would disappear in the modern world, ethnic groups appeared to provide benets to their members even in these changing contexts. In a famous article, Bates (1974) claims that ethnic groups would persist "because of their capacity to extract goods and services from the modern sector and thereby satisfy the demands of their members for the components of modernity". The competition generated by modernity is an incentive to form and maintain strong ethnic ties. These positions undoubtedly took the lead (Bentley 1987), and the key question became: what makes ethnicity be mobilized by individuals at a given time in a given place, rather than any other form of identication? The present paper stands in such a framework. Because of measurement problem, economists stayed out of this consequential evolution, and kept on tackling the role of ethnicity through the lens of indicators based on ethnic classications. In the 1990s, a wide literature emerged around the eects of ethnic diversity on development and growth, presenting fractionalization as one of the main reasons for social unrest, political instability or economic underdevelopment. The founding milestone is the article by Easterly and Levine (1997), which depicts the "African growth tragedy" as a consequence of the high level of ethnic heterogeneity. The vast majority of papers that followed use variously sophisticated fractionalization 2.

(4) indicators (see for example Bossert, D'Ambrosio, and La Ferrara (2006), Montalvo and Reynal-Querol (2005) or Esteban and Ray (1994)), taking fractionalization as an unvarying exogenous characteristic of a society and disregarding the variability of the salience of ethnic cleavages across time and space. Data availability of course explains why economists used these indicators rather than direct measures of ethnic salience. Until recently, no other measurement could be used than those based on classications, as the subjective and evolutive dimensions of ethnicity are much more complicated to capture quantitatively. The poor adequacy of such objective measures is all the more striking as these studies consider the subjectiveness of individuals as the driving factor for the weak social cohesion. Indeed, because it impacts the individuals' more or less intense feeling of trust (Alesina and La Ferrara 2000) and of membership to the community, heterogeneity is said to harm the quality of public policies (Alesina, Baqir, and Easterly 1999), the social and political stability (Easterly 2001), the quality of governance (Easterly, Ritzen, and Woolcock 2006), the level of investment (Montalvo and Reynal-Querol 2005) or infrastructures (Alesina et al. 2003), the prevalence of corruption (Olken 2006) etc. Fractionalization would lower the degree of empathy or trust that prevails in the society and therefore weaken the collective claim as well as the ability to dene common goals or set up authority and solidarity mechanisms necessary to the implementation of public policies. A recent series of works fortunately began to ll this gap. First, some papers focused on restricted areas to better disentangle the interactions between politics and ethnic salience. They showed convincingly how local politics could determine ethnic salience (Miguel 2004) (Posner 2004b) and how heterogeneity entails specic local equilibria (Miguel and Gugerty 2004) (Banerjee and Somanathan 2007). As a second important improvement, the lack of appropriate measures was recently lled by a set of surveys - most of them realized in the frame of the Afrobarometer project - which include some data on subjective identication and on the salience of ethnicity in the political eld. In this paper, we use an original set of surveys designed to analyze the informal sector in seven West-African main cities. These household surveys include "classical" information such as education, living conditions, health etc., but also several subjective questions on politics, governance or poverty. Among them is a question on the individual's ethnic identication, which made our study possible. We will come back to these data in section 2. With this new material, it is now possible to seek what determines the relevance of ethnic boundaries and the strength of ethnic ties. To the best of our knowledge, the political competition is the only explanatory dimension that has been investigated so far (Eifert, Miguel, and Posner 2007) (Norris and Mattes 2003). This approach focuses on the role of elites which may manipulate or at least exploit the feelings of ethnic loyalty and rivalry to achieve their goals (Posner 2006). It concurs with a variety of political science studies dedicated to the relationships between ethnicity and politics (Brass 1976) (Salamone 1985). In this top-down setting, the elites' quest for power appears as the driving force that fosters unity and 3.

(5) mobilization of an ethnic group. This kind of studies also includes the attempts to draw classications of "politically accurate ethnic groups" (see for example (Posner 2004a) or (Chandra 2005)) on the basis of a huge eort to compile information and evidence on the level of ethnic salience. The focus is put on the collective determinants of the mobilization of an ethnic group. Although we test some of these results in this paper (see Appendix 5), it is not our main axis. We rather focus here on individual incentives or determinants of ethnic identication. We must explicit here the conception of the nature of ethnic groups that underpins and legitimates such an investigation. On the one hand, ethnic group is strongly ascriptive. Even if some constructivist ethnological studies sometimes underline how some ethnic groups were divided, unied or even created from scratch for political purposes (Amselle and M'Bokolo 1985), we believe it is not the most common case. Ethnic groups and families may rather be seen as two aspects of the same reality - kinship - at dierent scales. In Africa, ethnic groups are very often structured by a common origin - mythic or not -, lineage patterns, rules of transmission and sophisticated genealogical structures. The griots (traditional singers and storytellers) carry out the function of genealogists, perpetuate the memory of extended families and contribute to cement ethnic groups by keeping alive the consciousness of common membership. An ethnic group may well be viewed as a set of embedded familial structures, and ethnic ties are best thought about in terms of kinship. It is thus sensible to assume that each individual is assigned to a given ethnic group at birth. But even if individuals know which is their ethnic group, ethnicity may be either mobilized or neglected voluntarily by the individual. This is the property of what some anthropologists called the duality of ethnicity (Van den Berghe 1976): it is both primordial to a certain extent - ethnic ties are innate, carry a duty of solidarity and reciprocity (Geertz 1963) and often have a sacred dimension (Shils 1957) - and it is however one of the many possible dimensions of identity that an individual can put forward. Fearon (1999) dened identity as the specic way an individual positions himself in the multiple social stratications at a given moment. Identity may then evolve according to the individual's interest or need (Fearon 2002), and more or less involve ethnicity. Ethnic identication may thus be subject to rational choices by individuals. Building on such a denition of ethnic identity, the relevant question becomes: what makes ethnicity be a facet of identity worth mobilizing by individuals? The microeconomics of social networks may provide many useful insights here. Social proximity - due to common ethnicity or to any other common characteristic - was shown to bring a lot to individuals in some specic social and economic contexts (Arrow 1972): solutions to agency problems in the enforceability of trade contracts (Greif 1993), formation of mutual insurance networks (Fafchamps 1992), intergenerational transmission of property (La Ferrara 2007), nancial decisions (Karlan 2007) or integration on the job market (Granovetter 1974). Ethnicity may be a powerful feature to build such social networks. As Habyarimana et al. (2006) have shown recently on the ground of 4.

(6) experimental methods, ethnicity may be used rationally as a "social focal point" that allows individuals to solve coordination problems, even if common membership did not create pre-existing social rules (Habyarimana et al. 2007). This provides individuals a strong rationale for deciding to take advantage of the ethnic membership. In this paper we investigate the individual conditions in which people tend to refer to their ethnic group1 . After having presented the data in section 2, we rst propose a general estimation of the individual determinants of ethnic identication (section 3). We show the impact of education and migration, as well as the intergenerational transmission of values. Our rst results on the impact of the job market lead us to investigate that dimension further, which we do in section 4. When individuals are willing to experience occupational mobility, they identify to their ethnic group if its insertion on the job market may help them reach a better job. We suggest in section 5 that these eects depend much on the fact that individuals have an initial deprivation of social capital, which makes ethnicity appear as a substitute. Section 6 concludes.. 2 Data In this paper we use a set of seven "1-2-3" surveys that were carried out by DIAL, AFRISTAT and the National Institutes of Statistics in the main cities of seven WestAfrican countries: Cotonou (Benin), Ouagadougou (Burkina Faso), Abidjan (Côte d'Ivoire), Bamako (Mali), Niamey (Niger), Dakar (Senegal) and Lome (Togo). These surveys are identical and therefore strictly comparable. They were carried on in the early 2000s on large samples representative of the urban population. The surveys all include the following question: "which group (or community) are you the most proud to belong to?". Two answers are possible: the ethnic group or the nation. The survey in Dakar is the only one which includes religion as a third possible answer. This may not be mixed up with ethnicity, all the more as 92% of the sample in Dakar declares being Muslim. We thus considered that religious identication may more accurately be aggregated to national identication. This leads to quite homogeneous rates of ethnic identication in the seven main cities (Figure 1). The way this question is formulated makes the answer disconnected from the arbitrary classications that were used before. When he answers the question, the respondent may think of any ethnic group or sub-group as his reference group. The survey thus does not impose individuals an ethnic identity that may be inappropriate, but leaves the door open to their own interpretation of which ethnic group they belong to. We then include a range of variables that describe the individual's situation in terms of education, occupation and migration status, along with classical demographical information. To study the alternative between national identication and ethnic identication, it makes sense to restrict the sample to adults - individuals older than 18 1 We. also examined some political factors that previous literature showed as also playing a role. The results are provided in Appendix 5. 5.

(7) Which group are you the most proud to belong to? % 90 80 70 60 50 40 30 20 10 0 Benin. Burkina Faso. Côte d'Ivoire. Mali. Nation. Niger. Senegal. Togo. Ethnic group. Figure 1: Frequency of ethnic identication in the main cities of seven countries. who have the nationality of the country considered. Our nal sample includes 30,042 individuals older than 18. Descriptive statistics are provided in Table 1.. 3 A rst estimation of ethnic identication We regress the binary variable "to be more proud of one's ethnic group" on the set of individual characteristics presented above. Our method is similar to the one used by Eifert, Miguel, and Posner (2007)2 . We use Logit estimations. Results are displayed in Table 2.. 3.1 The impact of education, migration and position on the job market We rst examine the eect of demographical or origin variables. Females appear to identify more to their ethnic group. This result holds when we add other observable 2 The. authors use the recently published Afrobarometer surveys on twelve mainly English-speaking East- or Southern African countries, and focus on the eect of the proximity of an election. They nd it to be a determinant of ethnic salience at the individual level. They also nd that ethnic identication is a positive function of the degree of competition that individuals experience in their quest for protected jobs.. 6.

(8) Benin. Burkina Faso. Côte d'Ivoire. Mali. Niger. Senegal. Togo. Total. Father farmer (%) Female (%) Age Size of the household. 36.2 51.6 34.5 5.8. 52.2 45.1 35.2 7.1. 47.8 49.3 31.9 6.8. 37.2 48.8 35.2 7.9. 47.7 50.8 35.0 8.0. 28.2 55.6 35.4 9.9. 40.7 46.3 35.5 4.7. 40.7 50.8 34.0 7.5. Born in the same place (%) Migrant from an urban area (%) Migrant from a rural area (%). 50.4 29.4 20.2. 43.6 43.8 12.6. 32.6 58.7 9.1. 48.1 35.0 16.9. 43.0 31.5 25.5. 60.9 28.5 10.7. 39.0 46.4 14.6. 47.63 35.81 16.56. Never attended school (%) At least attended primary school (%) At least attended secondary school (%) At least attended post-secondary school (%). 21.5. 42.1. 21.5. 48.0. 43.4. 35.2. 17.5. 30.5. 30.2. 22.8. 25.1. 15.3. 19.9. 33.4. 32.2. 26.1. 36.6. 28.1. 39.4. 30.1. 27.9. 25.3. 43.4. 33.4. 11.6. 6.9. 14.0. 6.6. 8.7. 6.2. 6.8. 10.0. Inactive excl. students (%) Student (%) Unemployed (%) Formal private sector (%) Informal private sector (%) Civil servant (%). 11.9 11.2 5.8 7.9 56.0 7.1 6,028. 12.2 6.2 13.4 7.4 49.6 9.6 1,642. 9.7 10.4 15.6 14.4 43.8 6.1 3,752. 17.7 8.2 9.9 8.2 48.1 7.5 4,399. 22.9 7.9 16.0 6.6 36.1 10.1 6,050. 22.3 6.4 15.9 8.8 41.7 4.6 6,668. 8.8 4.8 9.7 8.3 60.9 7.3 1,503. 15.1 8.7 13.4 10.3 45.6 6.6 30,042. Demographical background. Migration. Education. Occupation. Observations. Table 1: Descriptive statistics characteristics, meaning that it is not driven by the fact that being a woman is associated with some specic characteristics on education, migration, integration on the job market etc. Age appears to play a signicant negative role. This is robust to such changes as introducing age in a quadratic function, introducing a discrete variable of age classes or dropping the education variables that could interact with the eect of age in a context where schooling rapidly increases over generations in Africa. The eect of age may surprise: older people show a lower propensity to ethnic identication, although they are often thought to be more attached to traditional values while younger people are seen as vectors of modernity. The eect of the size of household also contradicts the view that there is a link between demographical transition and the loss of traditional values like ethnicity. Living in a more numerous household lowers the probability to refer to one's ethnicity, although the result does not seem very robust across specications and is no more signicant when country xed eects are added to the regression. We will come back to these surprising results on age and the size of households in section 5, after having investigated the existence of an active identication. We will see how they may be interpreted as revealing the fact that ethnic identication is a resource of social capital for those who are deprived of it. Having a father farmer (when the respondent is 15 years old) increases the probability to feel proud of one's ethnic group by 5% when this variable is introduced along with 7.

(9) (4) (5) (6) 0.03*** 0.03*** 0.02*** (0.05) (0.05) (0.04) Age -0.00*** -0.00*** -0.00** (0.00) (0.00) (0.00) Father farmer 0.02* 0.02* (0.05) (0.05) Size of the household -0.00* -0.00 -0.00*** (0.00) (0.00) (0.00) Migrant from a rural area 0.02 0.02** 0.03*** (0.07) (0.07) (0.06) Migrant from an urban area 0.04*** 0.04*** 0.04*** (0.05) (0.05) (0.04) Attended primary school -0.03*** -0.02** -0.04*** (0.06) (0.06) (0.05) Attended middle or secondary school -0.06*** -0.06*** -0.07*** (0.06) (0.06) (0.05) Attended post-secondary school -0.09*** -0.08*** -0.09*** (0.10) (0.10) (0.09) Inactive 0.01 0.02 0.01 (0.11) (0.12) (0.10) Student 0.02 0.02 0.02 (0.13) (0.13) (0.11) Unemployed 0.04* 0.04* 0.04** (0.12) (0.12) (0.10) Private informal sector 0.04** 0.04* 0.03* (0.10) (0.10) (0.09) Private formal sector 0.04* 0.04* 0.03* (0.12) (0.12) (0.10) Country xed eects no no no no yes no Observations 21596 20789 20733 20733 20733 27503 Log-likelihood -11685.62 -11252.04 -11165.39 -11158.26 -11049.93 -14650.52 Logit estimations, with marginal eect estimates (at mean values for the explanatory variables). Standard errors in parentheses. Signicantly dierent than zero at 90% (*), 95% (**), 99% (***) condence levels. Omitted migration status: "born in the city". Omitted education level: "never attended school". Omitted occupation: "civil servant". Female. (1) 0.04*** (0.04) -0.00** (0.00) 0.05*** (0.05) -0.00** (0.00). (2) 0.04*** (0.04) -0.00*** (0.00) 0.03*** (0.05) -0.00 (0.00) 0.03** (0.07) 0.04*** (0.05). (3) 0.02*** (0.05) -0.00*** (0.00) 0.02* (0.05) -0.00** (0.00) 0.02* (0.07) 0.04*** (0.05) -0.03*** (0.06) -0.06*** (0.06) -0.09*** (0.10). Table 2: A rst estimation of the individual determinants of ethnic identication demographical indicators only, but the magnitude and signicance of the direct eect are much more limited after controlling for the level of education. There may be a transmission of specic, more ethnic-oriented values to people with farming origin, but the main eect seems to be the intergenerational rigidity in schooling opportunities. We have shown that having a father farmer sharply decreases the probability to reach a high level of education in Africa (Bossuroy and Cogneau 2008). There may be both a direct and indirect eect here, and the fact that the signicance and magnitude of the direct eect sharply decreases when other variables are added suggests that the indirect eect is important. We then examine the impact of the migration status. Three categories may be constructed from the results of the survey: people born in the main city, people born in an other urban area (excluding the main city) and people born in a rural area. Table 1 8.

(10) shows that natives represent about a half of the population of the capitals, while urban migrants represent between 30 and 50% of the rest. Migrants from rural areas are the most numerous in Niger, where they represent more than one inhabitant of Niamey out of four. Being a migrant, wherever from, appears as a strong determinant for salience of ethnicity. A rst explanation may be that for cultural reasons, people born in the main city have a dierent set of values and feel more proud of their nation. We could expect that this is an eect of living in open, cosmopolitan cities as opposed to remote hamlets. But this explanation is weakened by the fact that being a migrant from an urban area is as good a predictor of ethnic pride as being a migrant from a rural area. This result could be due to the fact that we control for the father being a farmer, but it holds even when we do not control for this eect (column 6). An other more convincing explanation may thus be that being a migrant entails a specic attitude to ethnic pride, ethnic identication being mobilized as (or reecting) a strategy to get inserted in the society. Section 4 will provide evidence on this by analyzing the impact of ethnic groups' insertion on the urban job market, with a specic emphasis on the migrants' behavior. Education stands out as a very robust and signicant determinant of ethnic identication, with the highest magnitude: whatever the specication of the model, being educated has a strong negative eect on the propensity to refer to the ethnic group. All education modalities included in the regression have signicant negative impacts as compared to the omitted variable "having never attended school" (which concerns between 17.5% of the population in Lome, Togo, and 48.0% in Bamako, Mali). This nding is robust to shifts in the denition of education levels. Including the continuous variable "number of school years" instead of education classes also conrms the robustness of this result. Furthermore, the deterrent eect increases with the level attained: absolute values of marginal coecient estimates increase. When all other parameters are controlled for (column 5), having attended primary school lowers the probability to refer to the ethnic group by 2%, having attended secondary level lowers it by 6% and post-secondary education (a level reached by less than 10% of the sample) by more than 8%. There may rst be a direct eect of education on values and references. Education provides the knowledge and open-mindedness that allows people to refer more to their nation, and interactions with people of dierent backgrounds are frequent during education years, which may decrease the sense of ethnicity. But a high level of education also makes it less necessary to rely on ethnic solidarity to reach one's objectives, typically nd a good job. We will discuss this eect in the following sections. Including the individual's occupation status (column 4) reveals that the lower the level of integration on the job market, the higher the level of ethnic identication. With the most stable and protected occupational status - civil servant - taken as a reference, every modality appears to determine positively and signicantly the level of ethnic identication. Maybe we only capture the fact that civil servants have a certain ethos that makes them more prone to refer to the nation for which they work on a daily basis. But it is striking that being out of the job market (either student or any other kind of inactive status) does not have any signicantly impact on ethnic identication 9.

(11) when compared to being a civil servant. Inactivity is specic in that it is a situation where low competition prevails, since individuals are not (or no more) competing in the job market. Competition does hold among students, but being successful in the studies is not the same as being integrated on the job market: in principle, interpersonal connections and social networks are less necessary. So it may be that the level of precariousness that individuals experience on the job market determines their degree of ethnic identication, because of the will to experience an upward occupational mobility and the eciency of strategies based on ethnic networks. The insertion of the ethnic group itself should therefore be examined, which we do in the next sections. Interestingly -and consistently-, introducing a revenue variable does not help to explain the salience of ethnicity (results not shown). The status is the driver. The strategic dimension of ethnic identication for occupational mobility is tough to identify, since the eect is endogenous by nature: being in a precarious situation may lead people to strategically mobilize their ethnic membership, which in turn is supposed to improve their situation. However, one should note that this bias would cause the eects to be underestimated and not overestimated: since ethnic identication reduces the probability of, say, unemployment, the impact of unemployment on the propensity to refer to one's ethnic group should be even more important than what we observe. We thus may consider that the job market status appears to have an inuence, which we suspect to be related to the will of upward mobility, and we will investigate this strategic identication in greater details in section 4. But we rst turn to the role of the family.. 3.2 The role of the familial background We now try to investigate further the role of family background. In our initial estimation, we already introduced an occupation variable for the father, namely whether he was farmer at the time the respondent was 15 years old. The estimations show that having a father farmer increases the probability to refer to the ethnic group, more as a by-product of a low educational mobility than as the result of a specic intergenerational transmission of values. The surveys also include a question on the level of education of the father, but we have to take this variable with caution since the fairly low amount of missing values is very unequally distributed across the countries, with 43% in Senegal, 8% in Mali and not a single missing value in Burkina Faso or Côte d'Ivoire. The way it was coded is thus unclear and may reveal dierences between countries. Aside with these two variables, in the cases where the father and his child are members of the same household, we can also use the father's own identication group. The problem is of course that having the father and the child in the same household remember that we restrict our sample to people above 18 years old - does not occur very frequently, so that the number of missing values is very high. Besides, the observations selected by the availability of this variable are far from random for it requires that the child still lives at his father's place or that the father is hosted by one of his children. 10.

(12) We are likely to select younger people who have not yet founded their own household, individuals whose retired parents cannot provide for their own subsistence, or more traditional families (as compared to the urban standard) in which several generations live together. But keeping these caveats in mind, it may be interesting to focus on younger people of our sample and have a look on how identication passes over from generation to generation. We thus re-run our baseline regression on a sub-sample of young people (18 to 25 years old). Results are shown in Table 3. The sample size is smaller when the father's identication is introduced (Columns 1 and 4) because it requires that the father and the child live in the same household. Father proud of his ethnic group Father farmer. (1) 0.31*** (0.12). Father uneducated Female. (2) 0.04** (0.10). 0.01 0.03* (0.12) (0.08) Age -0.01** -0.01** (0.03) (0.02) Size of the household -0.00 -0.00* (0.01) (0.01) Migrant from a rural area -0.01 0.00 (0.30) (0.13) Migrant from an urban area 0.06** 0.04** (0.17) (0.10) Attended primary school -0.02 -0.03 (0.18) (0.11) Attended middle or secondary school -0.01 -0.02 (0.18) (0.12) Attended post-secondary school -0.02 -0.05 (0.32) (0.21) Inactive 0.13 0.05 (0.62) (0.37) Student 0.12 0.01 (0.60) (0.37) Unemployed 0.17 0.07 (0.61) (0.37) Private informal sector 0.16 0.06 (0.60) (0.36) Private formal sector 0.11 0.03 (0.65) (0.40) Observations 3356 6441 Log-likelihood -1503.83 -3447.31 Logit estimations. Marginal eects are presented. Coverage: 18-25 year old.. (3). 0.04** (0.08) 0.02 (0.08) -0.01*** (0.02) -0.00* (0.01) 0.02 (0.12) 0.05*** (0.09) -0.03* (0.10) -0.04** (0.11) -0.04 (0.20) 0.12 (0.39) 0.08 (0.38) 0.14* (0.38) 0.11 (0.38) 0.12 (0.41) 7437 -3910.82. Table 3: Family background and ethnic identication. 11. (4) 0.32*** (0.14) 0.04 (0.22) 0.02 (0.16) 0.02 (0.14) -0.01** (0.03) -0.00 (0.01) -0.01 (0.34) 0.09*** (0.20) -0.01 (0.21) 0.00 (0.22) 0.01 (0.36) 0.10 (0.73) 0.09 (0.70) 0.14 (0.72) 0.13 (0.71) 0.08 (0.78) 2445 -1106.09.

(13) The father's identication to his ethnic group appears as a strong predictor of the son's identication. The magnitude is high: if the father identies to his ethnic group, the probability that the child does the same is raised by more than 30%. The two other origin variables have the expected signs and are signicant. The level of education of the father plays a role even when the school attainment of the child is controlled for. When we combine the explanatory factors, the direct eect of value transmission overshadows the other ones. This is not too surprising, value transmission being a more direct channel and the composite product of many unobservable determinants (education by the family, socialization schemes, parental inuence...). To have a clear view on the individual determinants of ethnic identication, we may control for this background family eect. We introduce household xed eects in the regression, so as to highlight the determinants of individual's deviations from the household average. Results are presented in Table 7 in Appendix 1. This regression skims o and reveals the most powerful determinants of ethnic identication, those which signicantly make the individual deviate from his "natural" inherited reference group. These factors are migration from a rural area, high education and unemployment. The impact of being a woman obviously stays unaected by controlling for household xed eects. One more piece of evidence suggests that ethnic identication is not essential and permanent but may vary according to the individual's particular situation at a given moment. When the individual is married and the spouse(s) is present in the household, we can determine whether the marriage if exogamous or endogamous. Surprisingly, we observe no dierence in the level of ethnic identication between people who married a co-ethnic and people who married someone from an other ethnic group (Table 4). This shows how ethnic identication may be disconnected from the core cultural values that shape an individual's destiny and be more determined by the individual's particular situation and objectives. Endogamous marriage (77.0%). Exogamous marriage (23.0%). Is proud of his ethnic group (%) Is proud of his nation (%). 75.6 24.4. 75.0 25.0. Observations. 8 716. 2 781. Table 4: Exogamous marriage and ethnic identication This section thus highlights the importance of education and migration as key determinants of individual identication. But some determinants related to the individual's particular position at a given time seem to play a role. The insertion on the job market is a noticeable example, those being unemployed or in a poorly protected job being incited to identify to their ethnic group. The occupation status may foster a strategic 12.

(14) identication through the mobilization of kinship networks to improve the individual's position. This is the eect we now try to investigate more deeply.. 4 Ethnic identication and occupational mobility So far we have highlighted two main dimensions that drive ethnic identication. Our rst estimation revealed the role of education and migration, and we showed the inuence of the family background. But the results of the rst estimation also suggested that an other factor may be at stake concerning the position on the job market: the use of an "active" or strategic identication. In this section we try to investigate this point. We would have liked to show that identifying to the ethnic group actually improves the individual's well-being and that for a given initial situation, ethnic identication brings positive returns. This would have required to solve the inherent endogeneity problem which we brought up above. Not surprisingly, a naive regression - of revenue or job position on a vector of regressors including ethnic identication - reveals the opposite result. We tried to nd an instrument that would impact the level of ethnic identication but not the revenue or job position. Unfortunately, the surveys do not allow us to nd any convincing instrument. Though it would have been far from perfect, the best we could think of is the occurrence of an event that strengthens the ties within a community, like a birth or the death of a member of the group. In the expenditure part of the 123 surveys, the expenses for celebrations are mentioned as well as their purpose. But these data were collected several months after the rest of the survey used in this paper, so that we cannot use these expenses to predict the observed ethnic identication. We thus try to identify the active dimension of identication by showing that it is signicantly more used by the individuals whose characteristics make the most likely to implement these strategies. The channel through which the occupational position may determine the level of ethnic identication is the possibility for the individuals to try and take advantage of their ethnic belonging to make their way on the job market. People who do not have a sucient amount of human capital or social capital may play the ethnic card to nd a new job or have a promotion. This story has to do both with an initial deprivation of human capital and a will to experience occupational mobility. We thus take these two factors into account so as to reveal the active or strategic dimension of identication. The initial deprivation of human capital is largely captured by the individual's level of education. The will to experience mobility may be observed thanks to the fact that our surveys include a section on occupational perspectives for the individual. Any employed worker declares whether he intends to nd a new job in the same rm having a promotion -, to nd a new job in a dierent rm, or to keep the position he has at current time. In the rst two cases, the individual is asked in how many years he thinks he will obtain this new position. We dene people "eager to change job" as 13.

(15) those who want to change their position and intend to obtain it within a year. The other ones are said "non eager" either because they are not looking for mobility or because it is an abstract desire for which no strategy is likely to be implemented. All sample. 30.7. Female Male. 32.2 29.4. Non-educated 27.6 Father Primary 30.5 non-educated Middle / sec 33.5 Father Post-sec 31.7 educated Coverage: active people. 29.3 33.6. Unemployed Informal Formal Civil servant Father farmer Father non-farmer. 65.9 27.4 24.6 15.3 29.0 32.4. Table 5: Share of employed workers eager to nd/change job (%) The share of people eager to nd or change their job - around 30% of active people is quite homogenous across the sub-groups presented in Table 5, except for the subgroup of unemployed who of course are massively willing to change their situation. Not surprisingly, civil servants also stand out with a lower will to change their position. If we let aside these two exceptions, eagerness to move does not seem to reect any other observable characteristics. If we simply introduce this variable in the benchmark regression, it turns out being non signicant. The fact that ethnic identication may be used dierently according to the level of education of the individual may account for this. So we try to see if being looking for occupational mobility modies the way the job-market position aects the people's identication. As we explained above, this eect may dier according to the level of education. We test this possibility by introducing interaction terms in the regression. We provide the results in Table 8 in Appendix 2, but the number of civil servants who are uneducated and eager to change their job is much too low (23) for our results to be robust. They suggest that being unemployed or working in the informal sector signicantly raises the probability that an individual identies to his ethnic group provided that he is not educated and that he is willing to nd a job rapidly or to obtain a promotion. For those who don't long for occupational mobility, being unemployed or informal worker does not push them to identify to their group. Those who are educated have other ways to get a (better) job, so that their occupation status is not a key determinant. A poor insertion on the job market fosters ethnic identication for those people who want to improve their situation and are cornered to use this ethnic strategy because of a lack of human capital. A more robust strategy has to do with the insertion of the ethnic group itself. Since individuals may use their ethnic group as a stepping stone for reaching a favorable position in the society or enter the job market, their strategic implication in the ethnic group might depend on how far the group itself is well inserted. People will all the more ask for help in their job search as their co-ethnics actually are in position to do 14.

(16) something for them. We thus examine the eect of certain characteristics of the ethnic groups on the propensity of their members to identify with them. For doing this, we rely on the ethnic group classication used in the questionnaires which is quite similar to the ones used in the DHS national surveys. Of course, like any classication they are built upon somewhat arbitrary choices and methodologies. Table 9 in Appendix 3 presents the ethnic groups retained for each of the seven countries. A decisive parameter in the individual's ethnic identication may be the share of his ethnic group that is in a good position on the job market and could help him. The relevant group characteristic may vary from an individual to an other. For example, if identication is strategic, unemployed could all the more identify to their group as the employed share of this group is high: if their primary objective is to nd a job, this parameter will be the rst one considered. And they may be less inuenced by the share of ethnic group that is employed in the public sector. But while nding a job in the public sector appears as a second order aspiration for unemployed, it might be of rst order for informal workers while the share of employed people would be less important for them since they already are employed. We also saw in our core estimation in section 3 that migrants have a specic relationship to ethnic identication. This led us to also distinguish the migrants and people who were born in the city. If groups of migrants actually develop a specic solidarity, the migrant's strategic identication might depend on whether his ethnic group is represented by an important share of migrants in the city. The larger it will be, the more likely the individual would be to play the ethnic card. In order to test these hypotheses, we run regressions including the relevant shares of the ethnic group (one by one for they are embedded) in our baseline regression along with interaction terms involving the migration status, the will to nd or change job, and the occupational status. For robustness reasons, we only compute these characteristics on the 42 ethnic groups represented by at least 50 persons in the sample. Results are displayed in Table 6. The share of the ethnic group employed is a powerful incentive for people to identify to their ethnic group (rst column). We expected this eect to be higher for the people eager to nd or change their job. This is true only for the migrant people (the share of employed co-ethnics pushes the ethnic identication of the eager ones, and not of the non-eager). The distinction between eager and non-eager does not hold among the native. This pattern is clearer if we concentrate on unemployed people. The share of employed co-ethnics has a strong signicative eect on the ethnic identication of unemployed people if they declare being actively looking for a job. This holds both among migrants and natives. Among native people eager to change or nd a job, the share of co-ethnics working in the civil service (middle column) has a strong eect on the ethnic identication of employed workers, either informal or formal. But it has no eect on the unemployed, for whom the decisive variable is the share of employed people. For all people non eager to nd or change their job (migrant or native), the share of co-ethnics in the civil 15.

(17) Female Age Father farmer Size of the household Attended primary school Attended middle or secondary school Attended post -secondary school. Native Eager to change job Unemployed. Informal worker Formal worker. Non eager to change job Unemployed. Informal worker Formal worker. Migrant Eager to change job Unemployed. Informal worker Formal worker. Non eager to change job Unemployed. Informal worker Formal worker. 0.03*** (0.01) -0.00*** (0.00) 0.02* (0.01) 0.00 (0.00) -0.04*** (0.01) -0.06*** (0.01) -0.09*** (0.01). 0.03*** (0.01) -0.00*** (0.00) 0.02 (0.01) -0.00 (0.00) -0.03*** (0.01) -0.06*** (0.01) -0.09*** (0.01). 0.03*** (0.01) -0.00*** (0.00) 0.01 (0.01) -0.00 (0.00) -0.03*** (0.01) -0.07*** (0.01) -0.09*** (0.01). Share of the ethnic group employed. Share of the ethnic group in the civil service. Share of the ethnic group migrant and employed. 0.65** (0.01) 0.51** (0.01) -0.16 (0.77). 0.49 (0.64) 1.29* (0.08) 4.55*** (0.00). -0.00 (0.98) 0.21 (0.19) 0.27 (0.32). 0.29 (0.41) 0.42*** (0.00) 1.11*** (0.00). 0.63 (0.61) -0.36 (0.47) -2.07 (0.14). -0.14 (0.54) 0.31*** (0.00) 0.55* (0.07). 0.84*** (0.00) 0.45** (0.01) 0.94** (0.05). -2.34** (0.03) -0.33 (0.60) -2.26 (0.23). 0.53** (0.02) 0.43*** (0.00) 0.56 (0.12). 0.01 (0.98) 0.21* (0.05) 0.20 (0.49). -1.63 (0.24) -0.22 (0.62) -1.17 (0.26). 0.06 (0.83) 0.15 (0.11) 0.15 (0.49). Observations 18698 Logit estimations with interaction terms. Marginal eects are presented. Lower panel: P > χ2 in parentheses. Coverage: active people.. Table 6: Ethnic identication and group's insertion on the job market 16.

(18) service does not have any impact. The negative eect found for migrant unemployed people admittedly does not t this pattern. We nally turn to the share of co-ethnics who are employed migrants. This share has a strong positive eect on the people who themselves are migrant, either unemployed or working in the informal sector, and who are looking for a (better) job. All other categories are unaected by this variable, if we except one outlying coecient for the informal "non eager" native people. If we introduce the more general variable "share of co-ethnics who are migrants" (results not shown), there is no eect on ethnic identication. Though imperfectly, these results seem to sketch a fairly consistent picture. People tend to identify all the more to their ethnic group as a large share of their co-ethnics is in a good position to help them get a job or improve the job they already have. Migrants are responsive to the share of their co-ethnics who are also migrants and employed. This provides suggestive evidence on the existence of a strategic and active ethnic identication, echoing some sociological theories that emphasize the role of interpersonal ties in job search like the seminal work by Granovetter (1974). Our results suggest that his framework fully applies to the role of ethnicity for job search in African urban contexts, the "weak ties" of co-ethnicity being strategically mobilized to reach a more protected position. Recent economic studies have shown how social networks and kinship ties could be part of a deliberate strategy to integrate the job market (Luke and Munshi 2006). We go further by investigating how this may even involve the feeling of ethnic membership. The ndings on migrants also echo a branch of sociology on ethnicity and migration that was initiated by the American urban sociology of the 1920s. The Chicago School of Sociology typically studied thoroughly how ethnic groups would help the migrants to be integrated in American cities (see for example Louis Wirth's works on social integration of Jewish migrants in Chicago (Wirth 1927)). Our results concur to one of the most important results of these sociological works, namely that ethnic groups help their members getting integrated in the urban society and nd a place on the job market. Here also, recent empirical economic studies have highlighted the role of ethnic networks to help migrants integrate the job market (Munshi 2003). This section shed light on the existence of an active identication to the ethnic group as a strategy to improve one's situation on the job market. Individuals who want to have an occupational mobility are pushed to identify to their ethnic group by the precariousness of their job, and they do even more so as an important share of their group is positioned on the rung of the ladder just above the one where they stand, and may therefore help them climb it.. 5 Is ethnicity a resource in social capital? We saw how a poor education and being a migrant lead to a higher ethnic identication. We also studied how individuals are incited to identify to their ethnic group if a large 17.

(19) share of co-ethnics is in good position to provide help on the job market, this eect being all the stronger as individuals are less educated or migrant. These results appear to form a coherent picture if we consider the hypotheses that an initial deprivation of social capital fosters the use of ethnic identication as an investment in a specic form of social capital. In this short section, we discuss how ethnicity may be considered as a resource of social capital, in which people may invest when they are deprived of it and feel they need it. We refer to the notion of social capital as it is dened by Bourdieu as a measure of the individual's amount of relationships or size of networks (Bourdieu 1986) rather than to Putnam's denition of social capital as the set of norms, the level of trust, the density of interpersonal ties that one may observe at the level of a community or society3 (Putnam 1995). We try to identify what are the incentives for an individual to voluntarily invest in that kind of capital and accumulate it. By doing this, we refer to the works by Glaeser, Laibson, and Sacerdote (2002) who try to formalize individual investment in social capital. We built a small theoretical framework to illustrate how our results may concur to this view (see Appendix 4). Ethnic identication may be seen as a form of investment in a particular form of social capital: ethnic capital. It aims at extending an individual's network beyond the set of relationships integrated in an already accumulated stock of social capital. In this simple framework, the background of each individual is key for understanding its propensity to identify to it ethnic group. A person who enjoys a high initial level of social capital has little incentives to invest in ethnic relationships: since he already has an important network, the marginal benet he could derive from his implication in ethnic groups is limited, whereas its cost remains at the same level - getting involved in such networks is time-consuming and implies to be plausibly asked for reciprocity. This need to reciprocate is also the argument raised by Luke and Munshi (2006) to explain why high-ability individuals are less likely to get married: they do not really need marriage to get employed but would still bear the cost of high remittances to be sent to the extended family. A second key factor is the level of integration of the ethnic group in the job market. The nature of the initial stock of social capital is manifold. Two main dimensions were highlighted in the empirical estimations: education and local integration. Being native may help the individual have a better integration in the neighborhood and a better knowledge of local know-how (DiPasquale and Glaeser 1999), independently from his implication in his ethnic group. It also favors the creation and consolidation of a friendship network. Education gives access to information and integrates the individual in a network of other educated people. Education and local integration provide social capital, making ethnic identication less necessary. Conversely, ethnic identication is peak when the deprivation of social capital is maximum. Non educated people and uprooted migrants who arrive in the main city appear as the most likely to claim their belonging to the ethnic group. Moving to the city 3 For. a comprehensive survey on this notion, see Coleman (1990), Putnam (1995) and Sobel (2002). For a discussion on its interactions with development, see Woolcock (1998).. 18.

(20) engenders a shift in the reference group and a destruction of the stock of social capital: individuals leave their relationships and a familiar environment, and have to reconstruct a network in order to meet the challenges of arriving in the big city to earn a living. In an urban multiethnic environment, ethnicity becomes a distinctive element that helps people reconstitute the amount of social capital necessary for successful social interactions (Habyarimana et al. 2006). The intergenerational transmission of value may partly be the reection of the eect of the initial stock of social capital, which might indeed be inherited from the family background. The rigidity in intergenerational transmissions of occupations and education levels in the African context (Bossuroy and Cogneau 2008) accounts for the eects of the father's characteristics. This intergenerational rigidity may as well fully apply to the level of social capital (Bourdieu and Passeron 1979). The father's deprivation of social capital drives him to show a high level of ethnic identication, but also determines the child's own deprivation of social capital and thus his ethnic identication. This framework also helps understand the role of age. Time creates social capital by stabilizing individual's work trajectory and social group. Social integration consolidates with age and the competition on the job market decreases over the life-cycle, because the upward moves occur early in a work career (Bossuroy and Cogneau 2008) and because of the path-dependency of occupational trajectories. In sum, the deprivation of social capital from other sources decreases with age, while the need to mobilize this capital on the job market also decreases. This makes ethnic identication less necessary, and the salience of ethnicity decreases subsequently.. 6 Conclusion This study is a rst attempt to analyze which characteristics pushes an individual to identify to his ethnic group. We provide empirical evidence on the roles played by the level of education, the family background migration and the use of identication as an instrument for job search and occupational mobility. The two dimensions of identication - one being an inheritance of the educational and family background, the other being strategic - may however be parts of the same reality: in a society where a erce competition prevails, the initial deprivation of social capital due to a poor education or to migration creates the need to implement strategies based on ethnic ties to experience upward mobility. Since the intergenerational reproduction of social positions is high, this deprivation might be massively inherited. This chapter admittedly suers from shortcomings, mostly due to the complexity of the phenomenon under study. Pinning down the economic determinants and implications of such a subjective concept as ethnic pride turned out being much more complicated than expected. But we hope to have uncovered some suggestive results on the fact that ethnic salience is not only the product of the political manipulation of masses by the elite, but may also be a rational response of individuals facing such economic constraints as the lack of education or the diculty to get integrated on the job market. 19.

(21) Our results contribute to the analysis of the eects of modernity and development on identities. We provide empirical support to Bates' thrust that modernity raises the importance of ethnic groups because it stirs the competition between individuals and makes ethnic ties strategic. Our analysis also contributes to the sociological analysis of the interactions between an individual and his social group. The importance of intragroup solidarity reveals, especially among migrants. The destruction of social capital provoked by migration makes it often necessary for the individual to get involved in the community of migrant co-ethnics that would help him to make his way in the city of settlement. Identity choice therefore appears closely linked to some core characteristics of the economy. The rigidity on the job market and the returns to education appear as key parameters to understand why it is rational for individuals to cling to their ethnicity for improving their situation. Much has been said on politics being a determinant of the salience of ethnic cleavages. We suggest that the interaction may as well be analyzed the other way round. Social and economical contexts lead individuals to have a certain relationship to their ethnic group, and social rigidity causes the reproduction of these attitudes over generations. The impact of individual ethnic salience on politics is huge, for it determines the feeling of citizenship, the importance of patronage ties and the motives for voting. The capacity of the State to provide people with a uid job market that adequately rewards education partly determines the citizen's attitude to the nation. In this respect, the openness of a society, as measured by intergenerational mobility and open access to education and protected jobs, may appear as a pre-requisite for the construction of a nation and the settlement of democracy.. 20.

(22) References Alesina, A., R. Baqir, and W. Easterly. 1999. Public Goods and Ethnic Divisions. Quarterly Journal of Economics 114:124384. Alesina, A., A. Devleeschauwer, W. Easterly, S. Kurlat, and R. Wacziarg. 2003. Fractionalization. Journal of Economic Growth 8:155194. Alesina, A., and E. La Ferrara. 2000. The Determinants of Trust. NBER Working Papers, vol. 7621. Amselle, J.-L., and E. M'Bokolo. 1985. Au coeur de l'ethnie. Ethnicité, tribalisme et Etat en Afrique. La Découverte. Arrow, K. 1972. Gifts and Exchanges. Philosophy and Public Aairs 1:4:343362. Banerjee, A., and R. Somanathan. 2007. The Political Economy of Public Goods: Some Evidence from India. Journal of Development Economics 82 (2): 287314. Barth, F. 1969. Ethnic Groups and Boundaries. Edited by Brown. Boston. Bates, R. 1974. Ethnic Competition and Modernization in Contemporary Africa. Comparative Political Studies 6:4:457482. Bentley, G.C. 1987. Ethnicity and Practice. Comparative Studies in Society and History 29:1:2455. Bossert, W., C. D'Ambrosio, and E. La Ferrara. 2006. A Generalized Index of Fractionalization. IGIER Working Paper Series, vol. 313. Bossuroy, T., and D. Cogneau. 2008. Social Mobility and Colonial Legacy in Five African Countries. DIAL Working Paper 2008-10. Bourdieu, P. 1986. Chapter The Forms of Capital of Handbook of Theory and Research in the Sociology of Education, edited by John G. Richardson. Greenwald Press, Greenwald Press. Bourdieu, P., and J.C. Passeron. 1979. The Inheritors: French Students and their Relation to Culture. University of Chicago Press. Brass, P. 1976. Ethnicity and Nationality Formation. Ethnicity 3:225241. Chandra, K. 2005. Chapter A Constructivist Dataset on Ethnicity and Institutions of Identity as a Variable, edited by R. Abdelal, Y. Herrera, A.I. Johnston, and R. McDermott. Coleman, J.S. 1990. Foundations of Social Theory. Harvard University Press. DiPasquale, D., and E.L. Glaeser. 1999. Incentives and Social Capital: Are Homeowners Better Citizens? Journal of Urban Economics 45 (2): 354384 (March). Easterly, W. 2001. Can Institutions Resolve Ethnic Conict? Economic Development and Cultural Change 49 (4): 687706 (July). Easterly, W., and R. Levine. 1997. Africa's Growth Tragedy: Policies and Ethnic Divisions. Quarterly Journal of Economics 112 (4): 120350.. 21.

(23) Easterly, W., J. Ritzen, and M. Woolcock. 2006. Social Cohesion, Institutions, And Growth. Economics and Politics 18 (2): 103120 (07). Eifert, B., E. Miguel, and D.N. Posner. 2007. Political Sources of Ethnic Identication in Africa. Afrobarometer Working Paper. Esteban, J., and D. Ray. 1994. On the Measurement of Polarization. Econometrica 62 (4): 81951. Fafchamps, M. 1992. Solidarity Networks in Preindustrial Societies: Rational Peasants with a Moral Economy. Economic Development and Cultural Change 41 (1): 14774. Fearon, J.D. 1999. What is Identity (as we now use the word)? Working Paper, Stanford University. . 2002. Ethnic Structure and Cultural Diversity Around the World: a Crossnational Dataset on Ethnic Groups. Working Paper, Stanford University. Geertz, C. 1963. Chapter The Integrative Revolution: Primodial Sentiments and Civil Politics in the New States of Old Societies, New States, edited by The Free Press. New York. Glaeser, E.L., D. Laibson, and B. Sacerdote. 2002. An Economic Approach to Social Capital. Economic Journal 112 (483): 437458 (November). Glazer, N., and D.P. Moynihan. 1963. Beyond the Melting-Pot. Edited by Harvard University Press and MIT Press. Cambridge. Granovetter, M. S. 1974. Getting a Job. A Study of Contacts and Careers. University Press. Greif, A. 1993. Contract Enforceability and Economic Institutions in Early Trade: the Maghribi Traders' Coalition. American Economic Review 83 (3): 52548 (June). Habyarimana, J., M. Humphreys, D.N. Posner, and J. Weinstein. 2006. Social Focal Points. Working Paper. . 2007. Why are Co-Ethnics Believed to be More Trustworthy? Working Paper. Horowitz, D.L. 1985. Ethnic Groups in Conict. University of California Press. Karlan, D. 2007. Social Connections and Group Banking. Economic Journal 117:52 84. La Ferrara, E. 2007. Descent Rules and Strategic Transfers. Evidence from Matrilineal Groups in Ghana. C.E.P.R. Discussion Papers, vol. 6111. Leach, E. 1964. Response to Raoul Naroll's On Ethnic Unit Classication. Current Anthropology, vol. 5:299. Luke, N., and K. Munshi. 2006. New Roles for Marriage in Urban Africa: Kinship Networks and the Labor Market in Kenya. Review of Economics and Statistics 88(2):264282. 22.

(24) Miguel, E. 2004. Tribe or Nation? Nation Building and Public Goods in Kenya versus Tanzania. World Politics 56:32762. Miguel, E., and M.K. Gugerty. 2004. Ethnic Diversity, Social Sanctions and Public Goods in Kenya. Typescript. Montalvo, J.G., and M. Reynal-Querol. 2005. Ethnic Diversity and Economic Development. Journal of Development Economics 76 (2): 293323 (April). Munshi, K. 2003. Networks in the Modern Economy: Mexican Migrants in the U. S. Labor Market. Quarterly Journal of Economics, pp. 549599. Murdock, G.P. 1957. World Ethnographic Sample. American Anthropologist 59:664 687. Nadel, S.F. 1951. The Foundations of Social Anthropology. Cohen and West. Naroll, R. 1964. On Ethnic Unit Classication. Current Anthropology 5:283291. Norris, P., and R. Mattes. 2003. Does Ethnicity Determine Support for the Incumbent Party? KSG Working Paper Series, vol. RWP03-009. Olken, B. 2006. Corruption and the Costs of Redistribution: Micro Evidence from Indonesia. Journal of Public Economics 90(4-5):853870. Posner, D. 2006. Institutions and Ethnic Politics in Africa. Cambridge University Press. Posner, D. N. 2004a. Measuring Ethnic Fractionalization in Africa. American Journal of Political Science 48:849863. . 2004b. The Political Salience of Cultural Dierence: Why Chewas and Tumbukas Are Allies in Zambia and Adversaries in Malawi. American Political Science Review 98:529545. Putnam, R.D. 1995. Bowling Alone: America's Declining Social Capital. Journal of Democracy 6:1:6578. Salamone, F.A. 1985. Colonialism and the Emergence of Fulani Identity. Journal of Asian and African Studies 20:193202. Shils, E. 1957. Primordial, Personal, Sacred and Civil Ties. British Journal of Sociology 8:2:130145. Sobel, Joel. 2002. Can We Trust Social Capital? Journal of Economic Literature 40 (1): 139154. Van den Berghe, P.L. 1976. Ethnic Pluralism in Plural Societies: A Special Case? Ethnicity 3:242255. Wirth, L. 1927. The Ghetto. American Journal of Sociology 33(1):5771. Woolcock, M. 1998. Social Capital and Economic Development: Toward a Theoretical Synthesis and Policy Framework. Theory and Society 27:151208.. 23.

(25) 7 Appendix Appendix 1 (1) 0.31*** (0.05) Age -0.00 (0.00) Migrant from a rural area 0.24** (0.11) Migrant from an urban area 0.07 (0.08) Attended primary school -0.12 (0.07) Attended middle or secondary school -0.29*** (0.08) Attended post-secondary school -0.54*** (0.13) Inactive 0.10 (0.13) Student 0.15 (0.14) Unemployed 0.31** (0.13) Private informal sector 0.15 (0.12) Private formal sector 0.10 (0.14) Observations 7923 Log-likelihood -2897.55 Conditional Logit regression with household xed eects. Marginal eects are presented. Female. Table 7: Baseline estimation with household xed eects. 24.

(26) Appendix 2 Female Age Father farmer Size of the household Migrant from a rural area Migrant from an urban area Attended primary school Attended middle or secondary school Attended post-secondary school. Uneducated Eager to change job. Yes Yes. Unemployed. 0.54** (0.02) Private informal sector 0.52** (0.02) Private formal sector 0.48* (0.06) Logit estimations with interaction terms. Marginal eects are presented. Lower panel: P > χ2 in parentheses. Coverage: active people.. 0.02** (0.06) -0.00* (0.00) 0.01 (0.06) -0.00 (0.01) 0.01 (0.08) 0.04*** (0.06) -0.01 (0.32) -0.04 (0.32) -0.08* (0.05). Yes No. 0.07 (0.35) 0.06 (0.25) 0.10 (0.17). No Yes. 0.02 (0.80) 0.06 (0.17) 0.03 (0.60). Table 8: Eect of the position on the job market interacted with the level of education and the eagerness to change job. 25.

(27) Appendix 3 Benin. Fon Adja Yoruba Other Dendi Yoa/Lopka Bariba* Betamaribe* Fula*. Niger. Zerma Hausa Fula Tuareg Other Kanuri Gurma* Tubu* Arab*. 61.9 20.3 10.8 3.3 1.5 1.1 0.6 0.3 0.3. 52.0 32.0 7.4 5.0 1.6 1.4 0.5 0.1 0.1. Burkina Faso. Mossi Other Mandings Bissa Gurunsi* Other* Fula* Dagari/Lobi* Gurmanche* Bobo* Senufo*. Senegal. Wolof Fula Serer Lebou Jola Other Mandinka Sarakole Mandyak/Balanta. 80.8 4.4 4.1 3.0 1.9 1.8 1.3 1.2 1.1 0.5. 41.28 18.87 13.06 7.63 5.19 4.8 3.86 2.88 2.43. Côte d'Ivoire. Akan North Mande Kru Gur South Mande. Togo. Adja/Ewe Kabye/Tem Para/Gurma/Akan Ana/Ife* Akposso/Akebou* Other*. 45.3 19.2 18.9 9.7 6.9. Mali. Bambara Malinke Fula Sarakole Senufo Dogon Songhai Bobo Other Arab Tuareg*. 74.0 15.8 5.0 2.7 1.9 0.7. Table 9: Ethnic groups (% of the population in the capital) *: ethnic group represented by less than 50 individuals in the sample. 26. 35.1 18.1 17.0 11.4 4.2 4.0 3.6 3.1 1.9 1.3 0.3.

(28) Appendix 4: A small model of investment in ethnic capital Hypotheses We consider a single individual who belongs to an ethnic group and cannot change to another. His revenue or utility depends positively on his level of human capital, which is a combination of an initial stock of capital K0 ≥ 0 and a ux of social capital on which decisions are made. K0 typically depends on the individual's level of education, and is exogenous. The ux of social capital depends on the implication of individuals in ethnic networks, what we call investment in ethnic capital and note θ ≥ 0. It corresponds intuitively to the number of persons contacted for help on the ground of common membership, the time spent in meeting other members of the ethnic group, the energy spent on seeking information from the ethnic group. The returns to such an investment are determined by the extent to which the ethnic group is well inserted in the urban world, typically on the job market. S² is the share of the ethnic group ² that is in position to help the individual. This parameter is exogenous as well. In our framework, ethnicity is similar to a non rival public good. The returns to investment are not aected by any queuing eect. The level of human capital is given by 1. R(θ, K0 ) = (K0S² + θ)S². (1). . Note that the returns of investments in ethnic capital are positive and marginally 2 decreasing: ∂R ≥ 0, ∂∂θR2 ≤ 0. Imagine an individual who seeks an increase in social ∂θ capital and gets in touch with all the persons he can think of on the ground of common membership to the ethnic group. The rst person will presumably help him make the rst steps in the ethnic network, give him the basic information he has to know about the job market for instance, and may open many doors. If the individual continues to invest in ethnic capital and increase θ, the next person contacted may help but the bulk of the information would have been already transmitted. The tenth person might provide little returns, even if it happens that she is the one who provides the marginal surplus that helps the individual go beyond the threshold required for, say, get a job in the formal sector. It is thus sensible to assume that returns are marginally decreasing. Without any investment in ethnic capital (θ = 0), a person will have access to an amount of social capital corresponding to the initial stock accumulated K0 > 0: literacy, people met during the schooling years, access to information thanks to reading skills... Conversely, an individual who has never been to school can only rely on the returns to his ethnic investment to increase his level of social capital. We can see that K0 and θ are substitutes in a marginally decreasing production function. The higher K0 , the lower the returns to a given amount of θ: investment in ethnic capital 27.

(29) Figure 2: Returns of θ for a positive K0. only helps to reach out to the people who are not already part of the network initially accumulated. Resorting to ethnic investment however has costs: it is necessarily time-consuming and creates a duty of reciprocity. Being involved in social networks and taking advantage of them implies that the individual might be later asked to give back some of his time, information, contacts etc. Investing in ethnic capital creates a social bond with an obligation to reciprocate. The set of obligations created by the investment θ - its cost C(θ) - corresponds to the sum of a series of reimbursements, which we sensibly assume convergent to a nite limit c - an innite limit would imply an innite obligation to reimburse and thus deter any positive investment in ethnic capital. (2). C(θ) = cθ. We assume that the cost function is linear, contrary to the returns function: when the individual establishes a contact with any relationship or relative, he opens a form of drawing right for this person, no matter if he or she is the rst or tenth person contacted. Moreover, the time and energy spent in such an increase in ethnic capital have no reason to decrease. The liability created is the same for the individual, whatever the low level of returns granted.. An optimal level of investment in ethnic capital The individual's program is the following: 1. max U (θ) = R(θ, K0 ) − C(θ) = (K0S² + θ)S² − cθ The rst order condition determines the optimal level of investment in ethnic capital θ? : 28.

(30) θ? =.  1 1   ( Sc² ) 1−S² − K0S²  . S². if K0 ≤ ( Sc² ) 1−S² if K0 >. 0. (3). S. ² ( Sc² ) 1−S². The individual's utility is thus U (θ? ) = R(θ? , K0 ) − C(θ? ), and is given by  1 S² 1 ²  )( Sc² ) 1−S² if K0 ≤ ( Sc² ) 1−S²  cK0S² + c( 1−S S² U (θ? ) =  S²  K0 if K0 > ( Sc² ) 1−S². (4). It may be easily veried that U (θ? ) is a positive function of S² and K0 , and a negative function of c. The level of ethnic identication is positively impacted by the benets provided by the investment and negatively by its costs. A rise in K0 lowers the level of S² ethnic identication. Beyond a certain threshold ( Sc² ) 1−S² , investment in ethnic capital would only provide negative returns because of the too weak marginal gains compared to the costs. Individuals with a high initial stock of capital thus do not invest in their ethnic relationships. In terms of identication, these people would tend to claim less their belonging to the ethnic group, for they know they can rely on the high level of integration they already reached and thus prefer to avoid the costs attached to an investment in ethnic capital. This small model constitutes an illustrative framework which helps understand how our empirical results may form a consistent picture.. 29.